Select only those records where all linked records fulfill a condition

Question

I have a classical 1:n relation between two tables as given below:

I need to select only those customers where 'NOT OrderDate IS NULL' for all their orders. I started with

SELECT Customers.Id, Customers.LastName
  FROM Customers, Orders
 WHERE Customers.Id = Orders.CustomerId AND NOT Orders.OrderDate IS NULL AND ...

and wanted to build further with 'FOR ALL' but failed.

I tried the suggestions given in the answers but none gave the correct results.

Here is my workaround with two temp tables as shown below:

DECLARE @TableA TABLE (
    Id int,
    CountA   int
)

DECLARE @TableB TABLE (
    Id int,
    CountB   int
)

INSERT INTO @TableA (Id, CountA)
SELECT        Customers.Id, COUNT(Orders.Id)
FROM            Customers INNER JOIN
                         Orders ON Customers.Id = Orders.CustomerId
GROUP BY Customers.ID

INSERT INTO @TableB (Id, CountB)
SELECT        Customers.Id, COUNT(Orders.Id)
FROM            Customers INNER JOIN
                         Orders ON Customers.Id = Orders.CustomerId
WHERE (NOT Orders.OrderDate IS NULL)                     
GROUP BY Customers.ID

Select tA.Id 
FROM  @TableA tA INNER JOIN  @TableB tB on tA.Id = tB.Id
WHERE tA.CountA = tB.CountB

Both temp tables differ only in that respect that the first selects the group count without a condition in Orders and the second temp selects them with a condition. Then joining the two temp tables where CountA = CountB gives only those customers where all related Orders fulfill the condition.

If someone finds a more elegant way, please let me know.

Answer 1

Any suggestions how to tackle this?

In cases like this, you need to think not of finding the records where all the related records meet a condition.

Instead, think of finding all the records WHERE there does NOT EXIST a related record that breaks the condition.

Answer 2

The most straightforward way to write this query, is with ALL :

SELECT Customers.Id, Customers.LastName
FROM Customers
WHERE '2000-01-01' < ALL(SELECT OrderDate FROM Orders WHERE Orders.CustomerId = Customers.Id)

You could also write it as a group query on the orders table, something like

WITH CustomerOrderDateRange(CustomerId, MinOrderDate, MaxOrderDate) AS (
  SELECT CustomerId, MIN(OrderDate), MAX(OrderDate)
  FROM Orders
  GROUP BY CustomerId
)
SELECT Customers.Id, Customers.LastName
FROM Customers
JOIN CustomerOrderDateRange
  ON Customers.Id = CustomerOrderDateRange.CustomerId
WHERE
  CustomerOrderDateRange.MinOrderDate > '2000-01-01'

I think this is cleaner if you need multiple criteria, for example a max date range as well.

Answer 3

Just locate the records you want to exclude from it, and put in a not in clause

 select *
 from Customers
 where Customers.Id not in (
   SELECT Customers.Id
   FROM Customers 
   join Orders on Customers.Id = Orders.CustomerId
   WHERE Orders.OrderDate < '2000-01-01'
   group by Customers.Id 
 )